Re: ksoftirqd using mysteriously high amounts of CPU time

[Brad Laue]

Andrew Morton wrote:
> Please ensure that the machine was booted with `profile=1' on
> the kernel boot command line.  The cost of this is negligible.
>
> When the problem starts happening, run:
>
> sudo readprofile -r
> sleep 10
> sudo readprofile -n -v -m /boot/System.map | sort -n +2 | tail -40
>
> (make sure that /boot/System.map refers to the currently-running kernel)


Gave the machine a bit of a workout to create the problem sooner than 
usual, albeit on a smaller scale. ksoftirqd consumed about 4.7% of the 
CPU according to top after several concurrent wget operations, and had 
accumulated about 6 seconds of runtime, unheard of for a system uptime 
of 15 minutes; the output attached is during this.

Hopefully I'm right in assuming that; on all other systems I've 
experimented with ksoftirqd doesn't even flinch unless subjected to 
extremes.

Did some further testing with a machine of identical configuration 
(suffers the same issue too); removed iptables from the suspects list, 
as the issue exists sans those modules.

Hopefully the attached shows some irregularity. If not, I'll have to 
reply back in a few weeks when the problem recurs over the course of time.

Cheers,
Brad
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/